Hierarchical recovery of digital terrain models from single and multiple return lidar data.

摘要

A hierarchical terrain recovery approach for generating digital terrain models (DTM) from single and multiple returns lidar data is presented in this paper. The algorithm can intelligently discriminate between terrain and non-terrain points by using adaptive and robust filtering and interpolation techniques. It processes the image pyramid, bottom-up and top-down, to estimate high-quality terrain surfaces from lidar data with varying point densities and scene complexities. Using road and vegetation information, the algorithm is able to adaptively adjust thresholds to be suited to process changing contents in a large scene. The algorithms have been tested extensively using multiple medium- and high- resolution lidar datasets. The worst-case error is below 25 cm Linear Error (LE) 90 comparing the derived DTMs and the raw range images on bare surfaces when testing several lidar datasets.